Regression testing

After building the EMOD executable (Eradication.exe), it’s important to verify that Eradication.exe is performing properly. Regression testing is a method by which the built code is tested to see if it has “regressed” or moved backwards in any way, such as previously reported (and fixed) issues reappearing.

Within the EMOD Regression directory there are many subdirectories that correspond to different disease scenarios in a variety of locations. Each of these contains the configuration and campaign files needed to run the simulation and the reference output, which represents the expected results. These reference outputs have been calculated by the scientific researchers modeling each scenario.

EMOD regression testing consists of running one or more of these simulations and comparing the output to the reference output. If the two outputs are comparable, the test passes; if they differ, the test fails. Because EMOD is stochastic, a passing test will fall within some acceptable range of the reference output, rather than be an identical match. If a regression test fails, EMOD will produce a matplotlib chart of the first 16 channels in the InsetChart.json output report. You can then review the charts to identify the problem.

If you want to quickly compare a simulation output to the reference output, you can also run any of the regression scenarios as a typical simulation, as described in Run simulations. However, this will not include the comparison and pass/fail evaluation that conducts. In addition, if you choose to do this, be sure to specify a different output directory, such as “testing”, so as not to overwrite the reference output.